This section is divided into three subsections. The first explains the population, sample and variable dataset. The second subsection presents the criteria for selecting variables and data extraction. The last subsection deals with the models’ development.
3.1. Population, Sample and Dataset
In the Andalusian legislation [
45] regulating holiday rentals in Seville, these accommodations are designated as viviendas con fines turísticos (VFTs) (i.e., homes for tourism purposes). The legislation defines VFTs as “those situated in real estate located on land used for residential use, where accommodation services are offered for tourism purposes in exchange for payment of a regular price”. This law also states that VFTs can be rented in full (i.e., the entire flat) or in part (i.e., a spare room). In addition, the Andalusian Tourism Law [
46] states that different types of tourist accommodations, including VFTs, must be registered with the Registro de Turismo de Andalucía (RTA) (i.e., the Andalusian Tourism Registry).
As can be seen in
Figure 3, Seville registered an increase of 296.33% in the number of VFTs over the period analysed (i.e., January 2017 to March 2019). As the legislation that regulates VFTs (i.e., Decreto 28 2016) entered into force quite recently, no official records are available with which to analyse the growth of this type of accommodation over a longer period.
As
Figure 4 shows, VFTs already carry a similar weight to hotel establishments in relative terms. The rest of the accommodations comprise tourist apartments (TAs) (see
Figure 4), which consist of a unique registry of three or more apartments distributed in the same or different buildings. In some cases, TAs include common areas (as in hotel establishments) with a reception, lounge, restaurant and/or swimming pool.
Data for the study were based on Solano et al. [
50]. All VFTs registered as offering the full modality (i.e., the entire flat) in Seville were taken as a reference point for the total research population. The per room modality was excluded from the sample due to the distortion that could occur in the models if different services were compared. VFTs registered as offering the per room modality were also not considered significant because they account for only 7.5% of the research population [
49]. The total number of VFTs in Seville at the time of data collection (i.e., October 2018) was 3750 (see
Figure 3 above), with 3467 offering the full modality, from which a sample of 665 was extracted. The sample considered was the part of the population from which complete data relevant to the elaboration of the intended models could be obtained. Regarding data extraction, the RTA, which is ordered by registration number, was taken and, through a Google search with the following structure: “apartment registration no.” + “Booking” + “Seville”, the holiday rental profile in Booking.com was found. It was then verified if both the address of the accommodation and the registration number matched in RTA and Booking.com. This was followed by data extraction to a Microsoft Excel file, which was refined and later incorporated into SPSS statistics software. This final sample was selected from a total of 1623 cases because the VFTs included various offers with a different number of beds.
The variables adopted for the sample were based on the evidence found in previous studies analysed in the above literature review; these are listed in
Table 1. A descriptive analysis facilitated a fuller understanding of the sample. Thus, holiday rentals in Seville had an average price per day of approximately 162 euros (€) and were about a 15 min walk on average from the Plaza del Triunfo. With regard to size, the accommodations were on average 76 m
2 with four beds. Almost all the VFTs had a TV and washing machine, and approximately 40% of the sample had a balcony, terrace, patio or view. Only a small minority (3%) had a swimming pool.
3.2. Variables
The information obtained was extracted from Booking.com [
51]. The exceptions were the minutes needed to walk from accommodations to the spot of maximum tourist interest (MIN) (see
Table 1 above), which was calculated based on Google Maps [
52] and district index (DINDEX), which was taken from Tinsa [
53]. The VSAP variable’s score was estimated based on an evaluation carried out for the present study.
The price (PRICE) was extracted per accommodation and day for a two-day stay, which is the average stay in Seville according to the Centro de Datos Turísticos del Ayuntamiento de Sevilla [
54] (i.e., the Tourism Data Centre of Seville). This variable includes taxes and other added expenses (e.g., cleaning, when paid separately). During the data collection process, whenever a property offered different types of apartments at the same price, the one that offered the highest added value was chosen to reflect a rational consumer’s behaviour. Finally, the option for cancellation in a specific period and/or a partial refund was given priority. The non-refundable option was selected only when no other option was available.
Regarding the minutes needed to walk from accommodations to the spot of maximum tourist interest (MIN), Plaza del Triunfo was chosen for Seville because this square is located between the Cathedral of Seville and the Real Alcázar. These are the two most visited monuments according to the Centro de Datos Turísticos del Ayuntamiento de Sevilla [
54]. The present study hypothesised that this variable could affect prices negatively, since the shorter the time spent on reaching the spot of maximum interest, the more expensive the accommodations should become.
The district index (DINDEX) was constructed from the average price per m
2 of housing in Seville, according to the district in which the accommodations were located. The district with the highest price took the value of one, and the rest of the districts took a proportional value, as shown in
Table 2. This research hypothesised that a higher value per m
2 in the district implies a higher value of the property, which will be passed on to customers.
Additional tests were carried out on the models, taking each district as a “dummy” variable except for one, which served as the reference point. The inclusion of all the districts at once would have created a problem of exact multicollinearity in the models.
The accommodations’ size was reflected in the sample through two variables: the surface of the property measured in m2 (M2) and the number of beds offered (BEDS). This approach was based on the hypothesis that a larger-sized property implies a higher value, which will again be passed on to guests. Similarly, an offer that involves a higher number of beds implies a more significant expenditure on such things as electricity, water and cleaning, so this variable should positively influence prices.
The selection of the variables related to amenities and other accommodation features was carried out through a search on Booking.com [
51] regarding holiday rentals and tourist apartments (see
Table 3). The selection criterion discriminated between the most frequently present variables (e.g., WiFi and air conditioning) and the least common ones (e.g., a gym or spa), which were not considered in the models. The latter type of variables often has little impact on prices, so these amenities are not especially significant in price composition. Furthermore, less common variables can generate atypical cases that cause distortions in the models. The variables were thus selected from the remaining frequently offered amenities.
Regarding the view variable, only views including the city and/or emblematic monuments were included, thereby excluding views of courtyards and/or interior gardens. Concerning the availability of parking, both the establishments’ facilities and the presence of private parking in the surrounding area were considered.
Finally, the price was taken from various periods. Six intervals were considered (see
Table 1 above). The first four were high season, weekday (HSWD, price from 27 to 29 May 2019); high season, weekend (HSWE, price from 31 May 2019 to 2 June 2019); low season, weekday (LSWD, price from 14 to 16 January 2019); low season, weekend (LSWE, price from 18 to 20 January 2019). The last two price variables were from Holy Week, which is a special event in the city (SE1, price from 18 to 20 April 2019) and the April Fair, also an important event in Seville (SE2, price from 10 to 12 May 2019).
To estimate the seasonality variables’ weighting, the sample was divided into approximately two equal parts (see
Table 1 above) based on high and low seasons: high season was from April to September and low season was from October to March. A more significant weight was given to the high season due to the higher occupation rate during this period. A similar approach was used regarding weekends, as they represent more than two-sevenths of cases compared to weekdays, because of the marked increase in overnight stays on weekends. Given the much higher occupancy rate during Holy Week, the 2% that this week represents out of the total days of the year was quadrupled in weight. The same approach was used for the April Fair as the city’s second special event, so the weighting of the days involved was approximately double the normal 2% for this period.